The AI age. Key insights from Mary Meeker's landmark report

‘We've never had a market of five billion users that could be accessed so easily’ Mary Meeker

Mary Meeker, author of the iconic Internet Trends reports and founder of venture capital firm BOND Capital, has published a new study. 360 pages - about the present and future of artificial intelligence: trends, risks, economics, technology and geopolitics. This is not a review in the spirit of ‘what will happen next’, but a rather sharp and factual study.

While in the past big money was made on oil and railways, today capital is built on computing. Whoever controls GPUs and clouds controls opportunity. Monopolies have become technological again - only now it's at the level of chips and data centres.

We at Emat EOOD have read the primary source carefully. Below, without further ado, is the gist. What's changing? Who is winning? And why infrastructure now decides everything.

Unprecedented speed
Neither cloud technologies, nor social networks and mobile applications have penetrated business at such a speed. AI services enter the market in a matter of months and start making money almost immediately. AI-powered startups are reaching revenues in the tens of millions faster than at any time in history. And while the internet used to be the infrastructure - the foundation on which services and businesses were built, today it has become the backdrop. AI intelligence is following the same path, only faster.

Price and tokens
The speed at which AI ‘speaks,’ ‘writes,’ or ‘thinks’ has increased tenfold, while energy costs have dropped to almost zero. In two years, the cost of processing a million tokens has dropped by 99%. Now models can generate information quickly and cheaply. What used to cost hundreds of dollars now costs cents. New chips - like NVIDIA's Blackwell - use 105,000 times less energy for the same amount of computation than they did a decade ago. This has a direct impact on AI. The technology is getting cheaper, mainstream and embedded in infrastructure, products, interfaces and businesses.

Capitalisation of thinking
Meeker argues that the cost of training models like OpenAI's GPT or Google's Gemini is steadily increasing and measured in billions of dollars. At the same time, the emergence of cheaper models, local alternatives and open-source groups like China's DeepSeek are driving down the price and forcing a rethink of monetisation models.

Infrastructure and geopolitics
In this race, only those with data centres, stacks, channels and access to modern chips win, say Emat EOOD Bulgaria experts. Today, NVIDIA, Google, Amazon, Microsoft, Apple and Chinese tech giants are fighting for control over chips. They are actively developing their TPUs and Trainium, investing in cloud solutions and creating infrastructure for mass adoption of AI.

Integrated AI solutions
Technology moves faster than regulators. The advantage goes to those who quickly integrate AI into their business and production processes. That's why Emat company is already developing integrated AI solutions - from software and hardware development to real-time connectivity and processing of corporate data.

AGI: the technology race
Universal intelligence - AGI - does not yet exist, but its development has already become a strategic priority. The US and China have entered the technology race. The countries are starting to build infrastructure for something that doesn't exist yet. China is introducing AI technologies in defence, military medicine, logistics and autonomous control systems. The US is betting on computing and investing billions of dollars in its technology base.

The battle for techno-sovereignty
The US is building models, China is building ecosystems, investing in battlefield logistics, target recognition, cyber operations and autonomous decision-making platforms. DeepSeek, Qwen and dozens of other models are emerging at a phenomenal rate. And it's not about copying. China is reinventing the logic of development and it is no longer a competition between developers, but a struggle for techno-sovereignty.

Labour market: shift
+448% of job openings in AI in 7 years. Minus 9% in other IT segments. Companies are not just looking for programmers, but people who can work with new cognitive systems. The education market is growing. Online courses, microlearning, retraining - all this is no longer an alternative, but the basis of career sustainability.

A fracture, not a trend
Mary Meeker calls what is happening today ‘the new internet’. And it's not a hype, it's a paradigm shift. But, unlike the 90s, there is little time to swing. AI is not a separate industry. It's what's changing the architecture of everything. You can't wait any longer. Neither can playing demos.

In times of euphoria,‘ Meeker reminds us, ’it is important to remember a simple rule: invest only what you are prepared to lose, and don't put everything on one card.